Introduce the notion of "semantic result parameter". Handle differentiation of inouts via semantic result parameter abstraction. Do not consider non-wrt semantic result parameters as semantic results
Fixes#67174
The substituon map might contain proper specialization (e.g. t_0_0 -> Double) that is required to fully specify a derivate type (e.g. during reabstraction conversion from fully specified function type to a substituted one)
Fixes#65073
This essentially passes the members of a linear map tuple as individual arguments. It yields few nice simplifications:
* No linear map tuples at all for getters / setters
* No tuple formation / deconstruction around pullbacks
* Pullbacks with loops still use heap-allocated tuples
The changes are intentionally were made close to the original implementation w/o possible simplifications to ease the review
Fixes#63207, supersedes #63379 (and fixes#63234)
Single input variable might yield multiple adjoint buffers if control flow is involved. Therefore we cannot simply transfer debug info from the input variable: it will be invalid as we will end with multiple locations for a single "source" variable, and, even worse, might end with conflicting debug info as different buffers might be optimized
differently.
We do:
- Drop input argument number. This must be unique and we're not
- Correct variable name
lit.py currently allows any substring of `target_triple` to be used as a
feature in REQUIRES/UNSUPPORTED/XFAIL. This results in various forms of
the OS spread across the tests and is also somewhat confusing since they
aren't actually listed in the available features.
Modify all OS-related features to use the `OS=` version that Swift adds
instead. We can later remove `config.target_triple` so that these don't
the non-OS versions don't work in the first place.
`salvageDebugInfo` is called during SIL Mem2Reg and could produce misleading debug info in the following case:
```
%a = alloc_stack $MyModel.TangentVector, var, name "self", argno 1, implicit, loc "debug2.swift":37:17 ...
...
store %b to %a : $*MyModel.TangentVector
```
Such SIL could be created as a result of inlining (where store comes from the inlined function).
Before this patch we'd end with `debug_value` instruction with variable information, but without or incorrect location.
This caused LLVM IR debug info verifier assertions when there might be another instruction with complete debug info (including location) for the same argument.
After this patch we always reuse it from destination alloca
Fixes#58660
In order to transfer debug info from `debug_value` to `alloc_stack` we need to drop `op_ref` debug expression, as the latter instruction represents the location and not the value itself. Also, fix the debug info verifier as variable type was deduced from `alloc_stack` / `alloc_box` improperly: pointer type of instruction itself was used instead of underlying object type.
The issue is only exposed when both optimizations (SIL mem2reg at least) and debug info are enabled
Resolves SR-15849
Apparently, the parameter index calculation in #37861 was not always correct in presence of other pullback parameters (e.g. captures and non-differentiated args). Collect all inout parameters and collect inout parameter adjoints correctly.
Resolves SR-15891
AutoDiff/compiler_crashers_fixed/rdar71191415-nested-differentiation-of-extension-method-optimized.swift is starting to fail recently. Tracked by rdar://89055298.
- Introduce a workaround while dealing with getLoweredParameterIndices() results:
it operates on AST and therefore does not take into account captured arguments
which are "on side" on AST and explicit in SIL
- Ensure captured arguments (@inout_aliased) are handled in a same way as ordinary
@inout arguments while generating a pullback type
Fixes SR-15205
In `TangentBuilder` temporary allocations are emitted in order to make generic calls to `AdditiveArithmetic.zero` and `AdditiveArithmetic.+`. When we perform accumulation of two adjoint values that correspond to a lexical value in the original function, the stack allocations require a debug variable, as we use the original instruction's location. However such a debug variable wouldn't really make sense as these temporary allocations do not represent the original variable's adjoint. With lexical borrow scopes, we hit a crasher when creating temporary allocations when accumulating adjoints for lexical `begin_borrow` instructions.
This PR makes `TangentBuilder`'s temporary allocations use an auto-generated location. The existing debug info emission behavior is not changed, as we still emit `debug_value` when materializing an `AdjointValue` after any necessary adjoint accumulation.
When SILGenWitnessTable creates a decl ref for the witness of a derivative function requirement, it is using the requirement's derivative generic signature in the resulting witness decl ref. This is wrong because the witness may have a different derivative generic signature than the requirement, leading to a crash. This bug was never discovered because GSB's dark magic made it "just work", until requirement machine.
The fix is to store the matched witness derivative generic signature in `Witness` during type checking, and during witness table generation, use the witness' generic signature to create a witness decl ref.
Resolves rdar://84716758, rdar://84213107 and rdar://84987079.
1. When calculating the differential type of an original function with an inout parameter and when the inout parameter has a type parameter, the inout parameter should get a generic parameter in the subst generic signature of the differential but it currently doesn't. This causes SILGen to attempt to reabstract the differential value in the JVP protocol witness thunk, whilst the generic signature is lacking requirements, leading to a requirement machine error. This patch fixes the calculation so that the JVP's result type (the differential type) always matches the witness thunk's result type.
Wrong type:
```swift
sil private [transparent] [thunk] [ossa] @... <τ_0_0 where τ_0_0 : Differentiable> (...) -> @owned @callee_guaranteed @substituted <τ_0_0, τ_0_1> (@in_guaranteed τ_0_0) -> @out τ_0_1 for <τ_0_0.TangentVector, τ_0_0.TangentVector> {
%6 = differentiable_function_extract [jvp] %5 : $@differentiable(reverse) @convention(method) <τ_0_0 where τ_0_0 : Differentiable> (@in_guaranteed τ_0_0, @noDerivative @inout τ_0_0, @noDerivative SR_13305_Struct) -> () // user: %7
HERE ====> %7 = apply %6<τ_0_0>(%0, %1, %3) : $@convention(method) <τ_0_0 where τ_0_0 : Differentiable> (@in_guaranteed τ_0_0, @inout τ_0_0, SR_13305_Struct) -> @owned @callee_guaranteed @substituted <τ_0_0> (@in_guaranteed τ_0_0) -> @out τ_0_0 for <τ_0_0.TangentVector>
```
Should be:
```swift
%7 = apply %6<τ_0_0>(%0, %1, %3) : $@convention(method) <τ_0_0 where τ_0_0 : Differentiable> (@in_guaranteed τ_0_0, @inout τ_0_0, SR_13305_Struct) -> @owned @callee_guaranteed @substituted <τ_0_0, τ_0_1> (@in_guaranteed τ_0_0) -> @out τ_0_1 for <τ_0_0.TangentVector, τ_0_0.TangentVector>
```
2. `TypeConverter::makeConstantInterfaceType` is not passing down the derivative generic signature to `SILFunctionType::getAutoDiffDerivativeFunctionType` for class methods, and this was caught by RequirementMachine during vtable emission. This patch fixes that.
Partially resolves rdar://82549134. The only remaining tests that require `-requirement-machine=off` are SILOptimizer/semantic_member_accessors_sil.swift and SILOptimizer/differentiation_diagnostics.swift which I will fix next. Then I'll do a proper fix for workaround #39416.
The SIL type lowering logic for AutoDiff gets the substituted generic signature
mixed up with the invocation generic signature, so it tries to ask questions
about DependentMemberTypes in a signature with no requirements. This triggers
assertions when the requirement machine is enabled.
Disable the requirement machine until this is fixed.
